Estimation of glacier surface motion by robust phase correlation and point like features of SAR intensity images
Introduction
Glaciers act as important components in albedo feedback mechanisms of the local and global climate system (Rees, 2006). Glaciers also play vital roles as crucial reservoirs of global fresh water. For instance, some climatic threshold regions (e.g. Central Asia and India) characterized by water stress and densely population are reliant at a large extent on snow and glacier melt water (Pellikka and Rees, 2010). However, in the context of global warming, the melting glaciers are expected to make increasing contributes to global sea level rise (Gardner et al., 2013). It is estimated that melting glaciers account for approximately 29% of the observed sea level rise (Schubert et al., 2013). Therefore, researches on the behavior of glaciers and the glacier mass balance with geometry changes have drawn growing attention in the past decade. The monitoring of glacier surface motion as well as its velocity is an essential method to obtain the quantitative changes of glacier providing numerical clues to deducing past glacier changes and predicting future developments (Schubert et al., 2013).
The earliest monitoring and studies of glaciers depended upon the compulsory repeated field-survey by glaciologists. The researchers had to travel to far-flung glaciers, normally located in high mountainous areas characterized by hazardous environment, to collect field data only in few dozen points accompanying with time-consuming and expensive field seasons. Accordingly, due to the inaccessibility of many glacier areas and the environmental hazards, the acquisition of reliable measuring data in large scale was always a challenging task (Schubert et al., 2013). Remotely sensed data, especially the space-borne data, providing high resolution, trustworthy, large scale, and time-sequential measurements in a cost effective manner, is expected to be competent for the glacier monitoring. Among various remote sensing techniques, SAR features the abilities of ignoring weather condition and providing images all year long with both intensity and phase information and is becoming more popular for measuring the glacier motion.
In this paper, we present an approach for the estimation of glacier surface motion using repeat-pass SAR intensity images via point like features (PLF) and a robust phase correlation (PC) algorithm. We have chosen as site for our case study the Taku glacier in Juneau ice field located in southeast Alaska. The work consists of four main steps: the preprocessing of SAR intensity imagery, the selection of PLF, the dense matching using robust PC and the estimation of 3D motion of the glacier. The main contributions which are specific to measuring the glacier surface motion can be stated as:
- •
A dense matching method based on point-like features in SAR intensity images is developed for a robust workflow for estimating the dense glacier motion map.
- •
An improved robust PC algorithm combing the random sample consensus (RANSAC) algorithm with singular value decomposition (SVD) is introduced for matching SAR images, which can cope with strong speckles and noise while keeping a good matching accuracy between small patches.
- •
An adaptive refined version of Lee filter is developed for the despeckling of SAR images, while the sharpness and details of image patterns can be preserved.
- •
The feasibility and superiority of proposed approach motion from high resolution SAR are demonstrated and endorsed by both synthetic and real datasets.
This paper is structured as follows. Section 2 introduces related works in the field of glacier monitoring using spaceborne SAR data as well as the state of art of the phase correlation algorithms. Current research dealing with the Taku glacier is also introduced in this section. In Section 3 the proposed phase correlation algorithm is presented and the processing steps to measure the glacier motion are described in detail. In Section 4, the experimental results using synthetic data and real SAR data of Taku glacier are presented and analyzed. Finally, conclusions are drawn and future perspectives are outlined in Section 5.
Section snippets
Glacier motion estimation using SAR image
The optical imagery acquired by earth observation satellites is the most commonly used dataset for glacier monitoring, featured for its high resolution, wide coverage, lower cost and shorter revisiting period than traditional method such as field surveying. Many researchers have already used optical imagery for measuring surface displacements (Schubert et al., 2013), glacier topography (Fallourd et al., 2011), and velocity (Berthier et al., 2005). However, for the polar and alpine regions, the
Workflow
An overall workflow of the proposed glacier motion estimation approach is shown in Fig. 2. The workflow consists of three main phases: pre-processing of SAR intensity image (I), dense matching of an image pair (II), and estimation of glacier surface motion (III).
In the pre-processing phase, the SAR image pair is filtered by the proposed adaptive refined Lee (ARLee) filter, so as to simultaneously reduce the speckle noise and preserve the details of texture. External DEM data are introduced for
Experiments
Both simulated and real SAR intensity image data are used to assess the performance of proposed method and to measure the real glacier motion. In simulated case, synthetic SAR intensity images are generated from original optical images with added noises. In actual application case, the TSX intensity image pairs covering Taku glacier is employed, aiming to measuring the motion of glacier surface within a time period of 11 days. In addition, in the context of glacier velocity monitoring, one of
Image filtering result
In Fig. 13, the filtering results using different filters are exemplified. Besides the proposed ARLee filter, Boxcar filter, local sigma filter, Lee filter, enhanced Lee filter and non-local mean filter are employed as comparisons. The simulated SAR intensity image is generated with a noise variance of 0.3 and noise mean of 1. The window size of filters used is 7 × 7 (for the ARLee filter, it is the minimum window size). SMPI and EPI, aforementioned evaluation criterions for the performance of
Summary and conclusion
In this paper, the proposed approaches for estimating glacier surface motion in Taku glacier of Juneau icefield using repeat SAR intensity images via PLFs and the robust PC algorithm have been proposed. The proposed approach has been validated by the both simulated and real SAR data. The results obtained under a variety of test confirm the superiority and feasibility of the proposed approaches, especially for the high-resolution TSX images. Several conclusions can be drawn, as follows:
- •
The
Acknowledgments
We would like to thank Oliver Lang from Infoterra for providing us the TerraSAR-X data of the Juneau Icefield. Furthermore we thank Oliver Maksymiuk and Michael Schmitt from Photogrammetry and Remote Sensing, and Prof. Wunderlich from Geodesy, Technische Universität München for the kind directions and helps.
References (64)
- et al.
Surface motion of mountain glaciers derived from satellite optical imagery
Remote Sens. Environ.
(2005) - et al.
Ice velocity determined using conventional and multiple-aperture InSAR
Earth Planet. Sci. Lett.
(2011) - et al.
Evaluation of existing image matching methods for deriving glacier surface displacements globally from optical satellite imagery
Remote Sens. Environ.
(2012) - et al.
3-D movement mapping of the alpine glacier in Qinghai-Tibetan Plateau by integrating D-InSAR, MAI and Offset-Tracking: case study of the Dongkemadi Glacier
Global Planet. Change.
(2014) - et al.
Velocity estimation of glaciers with physically-based spatial regularization – experiments using satellite SAR intensity images
Remote Sens. Environ.
(2016) - et al.
Application of image crosscorrelation to the measurement of glacier velocity using satellite image data
Remote Sens. Environ.
(1992) - et al.
Glacier surface velocity estimation using repeat TerraSAR-X images: Wavelet-vs. correlation-based image matching
ISPRS J. Photo. Remote Sens.
(2013) - et al.
Potential and limits of InSAR data for building reconstruction in built-up areas
ISPRS J. Photo. Remote Sens.
(2003) - et al.
InSAR image co-registration using the Fourier-Mellin transform
Int. J. Remote Sens.
(2005) - et al.
Accuracy of differential shift estimation by correlation and split-bandwidth interferometry for wideband and delta-k SAR systems
IEEE Geosci. Remote Sens. Lett.
(2005)
Measuring two dimensional movements using a single InSAR pair
Geophys. Res Lett.
Glacier velocity monitoring by maximum likelihood texture tracking
IEEE Trans. Geosci. Remote Sens.
Glacier velocity estimation by means of a polarimetric similarity measure
IEEE Trans. Geosci. Remote Sens.
Monitoring temperate glacier displacement by multi-temporal TerraSAR-X images and continuous GPS measurements
IEEE J. Select. Top. Appl. Earth Observ. Remote Sens.
Permanent scatterers in SAR interferometry
IEEE Trans. Geosci. Remote Sens.
Extension of phase correlation to subpixel registration
IEEE Trans. Image Process.
Impact of resolution and radar penetration on glacier elevation changes computed from DEM differencing
J. Glaciol.
A reconciled estimate of glacier contributions to sea level rise: 2003 to 2009
Science
Coseismic horizontal offsets and fault-trace mapping using phase correlation of IRS satellite images: The 1999 Izmit (Turkey) earthquake
IEEE Trans. Geosci. Remote Sens.
Satellite Radar interferometry for monitoring ice-sheet motion—application to an Antarctic ice stream
Science
Velocities and flux of the Filchner ice shelf and its tributaries determined from speckle tracking interferometry
Can. J. Remote Sens.
Optical flow estimation using Fourier Mellin transform
A subspace identification extension to the phase correlation method [MRI application]
IEEE Trans. Med. Imag.
Measuring coseismic displacements with point-like targets offset tracking
IEEE Geosci. Remote Sens. Lett.
A method for sea ice thickness and concentration analysis based on SAR data and a thermodynamic model
The Cryosphere
Sea ice motion tracking from sequential dual-polarization RADARSAT-2 images
IEEE Trans. Geosci. Remote Sens.
The phase correlation image alignment method
Glacier surface velocity estimation using SAR interferometry technique applying ascending and descending passes in Himalayas
Int. J. Appl. Earth Observ. Geoinform.
Ice sheet motion and topography from radar interferometry
IEEE Trans. Geosci. Remote Sens.
Projecting global change impacts and sustainable land use and natural resources management in mountain biosphere reserves
Remote Sens. Mountain Environ.
Cited by (12)
Change detection of urban objects using 3D point clouds: A review
2023, ISPRS Journal of Photogrammetry and Remote SensingDetection of landslide induced by large earthquake using InSAR coherence techniques – Northwest Zagros, Iran
2020, Egyptian Journal of Remote Sensing and Space ScienceCitation Excerpt :Many scientists have utilized SAR data to detect the corresponding deformations caused by earthquakes (Ambrosi et al., 2018; Qu et al., 2017). SAR interferometry is a practicable tool for detecting land deformations caused by earthquakes(Chen et al., 2014; Clifton et al., 2002; Ganas et al., 2016), as well as for estimating glacier surface motion (Dai et al., 2016); via application of DInSAR techniques to measurement seasonal change (Fang et al., 2016). These studies showed that InSAR coherence contained relative information, which is vital in identifying different types of change on land surface.
Detecting urban changes using phase correlation and ℓ<inf>1</inf>-based sparse model for early disaster response: A case study of the 2018 Sulawesi Indonesia earthquake-tsunami
2020, Remote Sensing of EnvironmentCitation Excerpt :Correlation, an essential procedure behind the image registration process, has been recently applied to quantify 2D ground deformation (Rosu et al., 2015). Fang et al. (2016) used phase correlation in synthetic aperture radar (SAR) intensity images to estimate glacier surface motion. Three-dimensional coseismic displacements were computed from Lidar data in Moya et al. (2017).
Glacier Area Monitoring Based on Deep Learning and Multi-sources Data
2021, Advances in Intelligent Systems and Computing